Use a points-to analysis of def.getDef() here and see if there is an object whose creation node resides in read_datasets(). I don't think we need the direct relationship here as is currently done because the dataset case is a bit different from the previous case. In the previous case, we're only selecting tensor "generators," but in this case, we are looking for tensors coming out of datasets, whether their "new" or not. We're not currently at the point where we are tracking tensors through datasets.
The def is empty in the PA, because we don't have a class for datasets. We also don't have classes for tensors, so this is consistent with the other APIs.
Use a points-to analysis of
def.getDef()
here and see if there is an object whose creation node resides inread_datasets()
. I don't think we need the direct relationship here as is currently done because the dataset case is a bit different from the previous case. In the previous case, we're only selecting tensor "generators," but in this case, we are looking for tensors coming out of datasets, whether their "new" or not. We're not currently at the point where we are tracking tensors through datasets.https://github.com/wala/ML/blob/6e53776d319e732f9cdf894f99237de55fa92f23/com.ibm.wala.cast.python.ml/source/com/ibm/wala/cast/python/ml/client/PythonTensorAnalysisEngine.java#L119-L128